Bridged, layer-2 networks, such as Ethernet networks, provide services to areas where fiber optic lines do not extend and generally provide high data capacity at a low cost. A problem with bus and ring networks like the Ethernet is the possibility of a single point of failure causing the system to breakdown. A common solution is to design the network with redundant segments and loops so that there is more than one route to each node in a Synchronous Optical NETwork (SONET)-like approach (i.e., a layer-1 network). Redundancy and loops can, however, present another problem in which a broadcast packet or an unknown unicast packet results in a broadcast storm where each node receives and rebroadcasts the packet, causing potentially severe network congestion.
Current solutions for preventing single points of failures in ring networks, such as Ethernet Automatic Protection Switching (EAPS) systems, utilize an EAPS domain having a control virtual local area network (vlan) and at least one protected data vlan. The EAPS domain is associated with a master node linked to at least one transit node in a ring network.
An EAPS system operates in conjunction with the master node to detect a network failure by means of control messages sent between the nodes using the control vlan. During normal operation, the master node blocks the protected data vlan traffic from traversing its secondary port. During a network failure, the master node reroutes the protected data vlan traffic through its secondary port. When the network is restored, the EAPS system operates in conjunction with the affected transit node to prevent looping by blocking the protected data vlan traffic from traversing its restored ring port until notified by the master node that normal operation has resumed.
Current solutions, however, encounter scaling issues with respect to the number of protected vlans. For EAPS solutions, as the number of protected vlans increase, data traffic convergence time also increases across all vlans. This is because EAPS processes the protected vlans sequentially. What is needed is an adaptable process for selecting which vlans should be serviced first.